Modeling and prediction of powered parafoil unmanned aerial vehicle throttle and servo controls through artificial neural networks

Prashant Kumar,Bisheswar Choudhury,Amandeep Singh,Janakarajan Ramkumar,Deepu Philip,Ajoy Kanti Ghosh
DOI: https://doi.org/10.1139/dsa-2022-0040
2023-01-01
Drone Systems and Applications
Abstract:This study proposes a framework for developing a realistic model for throttle and servo control algorithms for a powered parafoil unmanned aerial vehicle (PPUAV) using artificial neural networks (ANNs). Two servo motors on an L-shaped platform, control and steer the PPUAV. Six degrees of freedom mathematical model of a dynamic parafoil system is built to test the technique's efficacy using a simulation in which disturbances mimic actual flights. A guiding law is then established, including the cross-track error and the line-of-sight approach. Furthermore, a path-following controller is constructed using the proportional-integral derivative, and a simulation platform was created to evaluate numerical data illustrating the route's validity following the technique. PPUAV was developed, built, and instrumented to collect real-time flight data to test the controller. These dynamic characteristics were sent into the ANN for training. A diverging-converging design was identified to obtain the best consistency between predicted and observed throttle and servo control values. For a comparable flight route, the control signal of the simulated model is compared with those of the actual and ANN-predicted models. The comparative findings show that the ANN-predicted and actual control inputs were almost identical, with an 80%–99% match. However, the simulated response showed deviation from the actual control input, with an accuracy of 50%–80%.
What problem does this paper attempt to address?